MIDA

Modellazione e Interrogazione di Basi di Dati Multimediali
(Modeling and Querying Multimedia Databases)

A CNR-coordinated project

A Multimedia Database Management System should support the storage of large amounts of data and the synchronized presentation of video, audio, etc., and also be able to deal with the effective and efficient retrieval of requested information based on the objects' content.

The purpose of the MIDA project is to provide theoretical and practical tools for Multimedia Databases modeling and querying.

A Brief Description of Scientific Objectives

Information retrieval from a Multimedia (MM) Database is a difficult task for which no satisfactory solution has been presented yet. The substantial difficulty in accessing MM objects by using their content is due to two main factors, different but tightly related:

  1. It is difficult to describe the "real" content of a MM object, given the different nature of the data, their intrinsic heterogeneity, and the complex relations between them. The inherent inaccuracy typical of every possible description of MM objects is such that common approaches for traditional databases are unsuitable.
  2. There is a gap between high-level domain-specific concepts and low-level features, that can be (semi-)automatically extracted from MM objects.

The general purpose of this project is to provide original contributions in order to solve above problems. Within MIDA, this objective will be pursued by dividing it in three interacting levels:

MM objects' content modeling
The general approach is at a formal level and the goal is the definition of flexible models able to deal with the inherent vagueness and heterogeneity of objects' content, taking into account the specific medium.
Query processing and access methods
The goals are the definition and the implementation of access methods able to support "similarity" queries - given a query object, find the most similar DB objects according to some features. Models for representation of correspondences between high-level concepts and MM objects are also considered, in order to efficiently exploit such correspondences during the querying phase.
Query languages
The purpose is the definition of a MM Database language, in order to support complex content-based query expressions. Such language should include features related to heterogeneous sources access and support the definition of hypermedia "views", allowing a customizable access to the database.

Partners

Research Unit Responsible
CSITE-CNR, Bologna Paolo Ciaccia
CNUCE-CNR, Pisa Fausto Rabitti
Dipartimento di Sistemi ed Informatica, University of Firenze Alberto del Bimbo
Electric Engineering Department, University of L'Aquila Paolino Di Felice
Department of Electronic Engineering and Information Sciences, Politecnico di Milano Franca Garzotto
Department of Engineering Sciences, University of Modena Sonia Bergamaschi
Istituto Policattedra della Facoltà di Scienze MM.FF.NN., University of Verona Letizia Tanca